While much of the maritime sector is just beginning to explore digital twins and predictive maintenance, Canarian naval engineer Fernando Martínez Marco has taken a step further by creating a virtual applied research laboratory. In this environment, he merges physical models, neural networks, knowledge graphs, and Bayesian networks to rethink the design, operation, and regulation of maritime transport. A new field emerges at the intersection of naval engineering and artificial intelligence, still little known outside specialized areas. Notable advancements include digital twins capable of anticipating the fatigue of underwater structures, deep learning models that replace physical sloshing tests, semantic search engines that interact with international maritime regulations, and Bayesian networks that optimize search and rescue operations. These innovations shape an emerging landscape in the industry. From his personal AI Research Lab, Martínez Marco openly publishes results in "Engineering Intelligence," a platform showcasing prototypes and technical studies aimed at a future where naval systems act as cognitive infrastructures. The lab is defined as an exploration space at the intersection of naval engineering and AI, where deep learning architectures, knowledge graphs, and retrieval-augmented generation (RAG) systems are tested on issues such as material fatigue, ship inspection, or the application of SOLAS and IMO regulations. SOLAS (Safety of Life at Sea), managed by the International Maritime Organization (IMO), sets minimum maritime safety standards. A highlighted example is the digital twin for monitoring fatigue in deep-water risers. In "Digital Twin for Riser Fatigue Monitoring," Martínez Marco proposes a methodology that combines minimal sensors and machine learning to estimate fatigue damage under extreme conditions, achieving errors of 5-10% compared to high-fidelity references and significantly reducing instrumentation costs. In the field of liquefied natural gas (LNG) vessels, the lab addresses both operation and preliminary design. The study "Emissions Reduction in LNG Shipping" demonstrates how predictive pressure models in tanks can reduce greenhouse gas emissions by up to 86% during challenging voyages. Meanwhile, "Taming Sloshing with Neural Networks" uses deep neural networks and large experimental databases to predict extreme sloshing loads, accelerating preliminary design and decreasing the need for costly physical tests. AI for safety and risk analysis A crucial part of Martínez Marco's work focuses on safety and quantitative risk assessment, where decisions must be made under uncertainty. In "Dynamic Risk Assessment in Offshore Platforms," he combines the DEMATEL method with Bayesian networks to create a Dynamic Risk Assessment (DQRA) framework that allows real-time risk updates, identifying dependencies and relationships between safety barriers that traditional approaches miss. This logic is also applied to grounding accidents in "Ship Grounding Risk Assessment," where an advanced platform based on dynamic Bayesian networks and Noisy-OR causal logic is used to model the evolution of damages, from structural breaches to loss of vessel stability. Maritime search and rescue The "Bayesian SAR Orchestrator" project applies this probabilistic approach to search and rescue (SAR) operations. Using a dynamic Bayesian network that integrates real-time atmospheric data and drift coefficients, the system generates probability heat maps that optimize the deployment of rescue resources, replacing generic patterns with data-driven strategies. Thus, AI enhances maritime safety by providing decision-makers with a more accurate probabilistic overview of the location of people in danger. From checklists to regulatory intelligence Another block of projects explores the boundary between engineering and maritime regulation. In "Graph-Based LNG Inspection Intelligence," Martínez Marco develops a Port State Control inspection support system based on knowledge graphs and Graph Convolutional Networks, along with natural language processing (RoBERTa), achieving 87% accuracy in predicting deficiencies and promoting dynamic risk-based inspections. In "AI Semantic Search for Maritime Regulations," he creates a question-and-answer engine for international maritime regulations based on RAG, achieving 94% accuracy in retrieving complex rules and allowing offline execution on conventional hardware thanks to model quantization. This semantic search engine acts as an interpreter of SOLAS and IMO regulations for engineers, operators, and inspectors. Domain-specific AI in the maritime sector The set of projects from the virtual laboratory covers the entire maritime value chain: from ship design and structural integrity to emissions reduction, operational safety, inspections, and automated interpretation of regulations. The "Technical Expertise" section reflects this dual specialization, combining competencies in LLMs, Graph-RAG, deep learning, and NLP with experience in gas vessels, SOLAS/IMO regulations, stability, and maritime operations. More than a generic AI, Fernando Martínez Marco's work exemplifies "domain-specific AI": statistical and machine learning models adapted to the languages, physics, and constraints of naval engineering. His virtual laboratory anticipates a new generation of tools where expert knowledge is encoded in graphs, neural networks, and probabilistic systems, paving the way for maritime infrastructures that are not only mechanical or digital but also cognitive.